topic Re: Looking for a PROC to compare summary mean and proportion in Statistical Procedures

Looking for a PROC to compare summary mean and proportion

Emma_at_SAS — Tue, 04 Aug 2020 02:10:02 GMT

Hi everyone,

I want to compare the means and proportions between two different samples. The two data sets do not have common individuals or variables. I need to compare some common measures between these two samples.

For example, comparing age in sample 1 vs. sample 2 and %male in sample 1 vs. sample 2.

Instead of merging the two data, I want to create the summary statistics for each sample and use them for my comparison.

Do you know any procedure in SAS that I can use to compare summary statistics?

Thank you!

Re: Looking for a PROC to compare summary mean and proportion

mkeintz — Tue, 04 Aug 2020 02:24:21 GMT

You say that

The two data sets do not have common individuals or variables.

If you don't have common variables, then how do you know the mean age, or proportion male in BOTH datasets. Do you merely mean that the varibles in the datasets have the same information (i.e. age in year, and two values for sex), but those variables don't have the same name (or in the case of sex, the same variable type)?

If so, then you could modify one of the datasets to emulate the variable attributes (e.g. varname, vartype) and coding scheme (e.g. "M", "F" instead of 0,1) in the other data set. Then you're ready to combine the compatible data sets and execute a t-test (age contrast) and a proc freq (for sex distribution comparison).

Re: Looking for a PROC to compare summary mean and proportion

Norman21 — Tue, 04 Aug 2020 05:16:08 GMT

If you don't have (or don't want to use) the raw data, and only have summary data, some coding is required. More here:

https://www.lexjansen.com/nesug/nesug06/an/da24.pdf

Re: Looking for a PROC to compare summary mean and proportion

PaigeMiller — Tue, 04 Aug 2020 11:52:59 GMT

Instead of merging the two data

(Perhaps you mean "instead of appending the two data")

This is a restriction which makes the programming much harder and ought to be eliminated. You (or your company or university) is paying for all the work already done by SAS to compare means and percents, plus SAS has tested it, fixed bugs and proved that these work in millions of real world applications — don't try to do this yourself.

Use the RAW data. Append the two data sets. This makes everything you do much simpler. Compare the means by PROC TTEST or PROC MEANS (or SURVEYMEANS). Compare the percents by using PROC FREQ (or SURVEYFREQ).

Re: Looking for a PROC to compare summary mean and proportion

Ksharp — Tue, 04 Aug 2020 11:50:02 GMT

If there are only two group, compare MEAN of two group by PROC TTEST or PROC N1PARM + willconx.

compare PROPORTION(a.k.a risk ratio) of two group by

PROC FREQ ;

TABLE group*gender/relrisk ;

run;

@Rick_SAS wrote bolgs about these before.